Continuous, best-in-class open source sleep classification with extreme runtimes

This research addresses the challenge of assessing fatigue risk in shipboard environments. The submitting team, Arcascope, is composed of sleep and circadian rhythms experts with extensive experience in running algorithms on consumer wearables, including Android Wear devices and the Apple Watch. Our family of algorithms, trained on gold standard data collected by our team and released open source, span a wide range of feature complexity, model sizes, and hardware computing needs. As such, they can run on a number of different devices. Because the input data is raw acceleration data (which can be used to infer breathing and heart rate patterns), our approach is device agnostic for any device that has a sufficiently high quality accelerometer.  In this work, we propose to adapt our most recent algorithms into an even more lightweight form that can run on Apple Watch and Android Wear. However, as such devices generally come nowhere close to the extreme long runtime needs of the DoD, we further propose to integrate our algorithms with an existing family of wearables, currently in use by the Navy: the Matrix PowerWatch.  By integrating the PowerWatch’s low-energy design and advanced energy harvesting technologies with submitting organization Arcascope’s best-in-class open-source algorithms for sleep, we will seamlessly and simultaneously address the Navy’s need for reliable algorithms and extreme long runtime devices. Moreover, by building out this capability on other consumer devices, such as Android Wear and Garmin watches, we will reduce the dependence of the Navy on a single device. The work will build towards incorporating high quality sleep classification into the existing Hub infrastructure, enabling massive scale up of sleep classification capabilities.  This work will significantly enhance military readiness by providing a reliable, open, long-lasting wearable technology solution for continuous sleep monitoring and fatigue assessment in challenging environments, towards the goal of continuous health, sleep, and activity monitoring of military personnel.